Re: [CAD_CAM_EDM_DRO] Re: Out of Estop.. BUT

Hello Jon,
This is getting back to what I was suggesting many month ago:

It is kinda silly to go from a digital command in the motion controller to
an analog output, then to process it in the analog form and adding the extra tach
signal and then finally convert it back to a digital PWM signal.

There has been several articles about encoder to tach conversions and
software/speed problems. If I remember right, in IEEE Industrial Electronics.
That was several/many years ago that I saw the articles. Today it ought to
be extremely easy to do with super fast DSPs cheaply available.

Ideally, we should do away with all the analog stuff and just output a digital
command, and then let the DSP read the encoder signals and convert them to
speed and finally calculate the PWM needed. All feedback and filtering
would be done in software and programmable without screwdrivers and pots.
Nice and clean!
Bertho Boman
===============================================

Jon Elson wrote:

> Ah, yes, instead of a DC tach, it uses the encoder for velocity
> feedback. The problem is it has no velocity info between encoder
> pulses, and the velocity info it does have gets 'bumpy' when the
> speed is extremely low. My servo system, with DC tachs, can go
> down to about .01 IPM before the stick-slip of the table overwhelms
> the tach signal, and it starts to bump. This is so slow, you really
> can't even see it, but you can read it off the position display.
> The only way to do better is to go to a finer resolution encoder.
>
> Matt Shaver and I tossed this problem about for some time, as
> he wanted me to design an encoder to tach converter. My read
> was that it would be VERY hard to do, and get both very fast
> update at high speed and freedom of fluctuations in the output
> at very low speeds. My servo system has meaningful bandwidth
> up to sevral KHz, and so the delay of the conversion would have
> to be kept very low. This is a killer digital filtering problem, and
> is probably totally insoluble any other way.
>
> Jon